Hello Monks,

My first encounter with DBIx::Class's subquery construct went so well (and my need of n subqueries so great) that I tried a programmatic way of creating nested subqueries, shared here.

Context is simply a tag search function using 3 tables. DB::TagDef (not shown) holds string to id, DB::Post has 0-n tags, and DB::TagCloud1 contains the join. Tag_ids are retrieved from DB::TagDef in ascending order of weight to ensure the smallest possible start set, getting progressively smaller.

sub recursive_subquery{
    my($schema,$join_column,$key,$values)  = @_;
    my ($value1) = shift(@$values);
    my $sub_query =$schema->search({
        $key => $value1,
    });

    for(@$values){
        my $tmp_query = $schema->search({
            $join_column => { 
                -in => $sub_query->get_column($join_column)->as_query 
            },
            $key => $_,
        });
        $sub_query = $tmp_query;
    }
    return $sub_query;
}
[download]

And a short example of using it manually in Catalyst:

$c->stash->{rows} = $c->model('DB::Post')->search({
    'posts.post_id' => {
        -in => recursive_subquery(
        $c->model('DB::TagCloud1'),
        ,'post_id',
        'tag_id',[
            16042,
            190712,
        ])->get_column('post_id')->as_query 
    },
});
[download]

As an aside, the performance of this highlighted the unhappy fact that mysql tends to evaluate sub queries from the outside-in, so you may want to reverse the weighting in that setup.